Statistical Analysis Exercises

Last update: 17 Mar 2023 [History] [Edit]

This is a summary of the statistical analysis exercises to be done using TRexFitter. More details are available in the introductory talk linked from the tutorial agenda and in the statistics tutorial repository.

You can access the tutorial material on gitlab.

The exercises you will perform are:

  1. Read through the tutorial and run the example leptoquark fit
    • Setup the environment and inspect the TRExFitter config file (use documentation for reference)
    • Run the fit and produce / inspect the common plots from the Asimov fit (pre-fit, post-fit, tables, correlation matrix, pull plot, gammas, normalization factors, ranking plot, …)
    • Unblind the analysis and re-run the fit and plots; compare to the results from the Asimov fit
  2. Using the plots from 1. demonstrate that Drell–Yan and Top background normalization cannot be constrained simultaneously with the default fit setup
    • Find a solution (e.g. only one normalization factor) and compare results
  3. Create TRExFitter config files for the remaining signal mass points (400, 700, 1800 GeV)
    • Run the fits / inspect the plots for all mass points
    • Calculate the observed and expected upper limit on signal strength for each mass point
    • Determine the lower limit on the LQ mass
    • Bonus: make a plot of all limits (via TRExFitter multi-fit functionality)
  4. Instead of fitting the LQ mass, fit a BDT distribution obtained yesterday
    • Compare the results to the mass fit
  5. Be creative and explore other capabilities of TRExFitter (e.g. likelihood scan, signal significance)